I have a map that is, more or less, an elaborate density representation of the number of times a state is referred to throughout my corpus of 157 American literary texts from 1890-1930. I took the liberty of representing these counts relative to an artificially predefined set of regions.
Given my penchant for debilitating uncertainty, I am not quite sure what to say about the results. As with several DH projects, my inclination is to say that I learned more in creating the map than from observing what is currently in jpeg form above. That being said, there were some results that I was happy to observe.
On the one hand, it excites me that Louisiana and Mississippi are referred to the most with respect to the southern region. This, in many ways, confirms the longstanding hypothesis that Cajun and bayou regional culture was a prominent geographic focus of American literary attention.
Furthermore, it excites me that Kentucky and Tennessee also register quite highly, as I am currently toying with the notion that Vanderbuilt’s rising prominence in the literary community would very much heighten the attention of this region.On an extrapolated register, I suspect that, as the twentieth century would progress, the general influence of colleges would be a major factor in geographical imagination.
On the other, the map fails in some spectacular ways and this is perhaps due to sample size.
For example, the fact that Nevada does not register in my corpus is misleading as the Las Vegas strip is often referred to as symbolic function of commercial built space. Even though my scripts account for Las Vegas as a reference to Nevada, my corpus does not contain such references.
Furthermore, I am suspicious of the prominence that Washington state has and that my collected data does not contain many references to the capital. For reasons elaborated in my previous post, I anticipate that references to Washington were allocated to the state when they should have been instead registered as a reference to the capital.
Despite the oversights inherent to the imprecision of my computational automation, I did find this to be a wonderful exercise in multiple capacities. This makes it sound like there is something erroneous with the available tools, and I’m quite willing not to take all the blame, but I accept that I still have quite a bit to learn computationally and cartographically.
This is perhaps why, on a more utility-oriented note, I enjoyed this project. To a certain extent, I did learn more about developing functions for text-mining and representing those results in ArcGIS. It was an exercise primarily about translating actual textual data into geographic registers in a way that literary scholars are not prone to.
On a more conceptual level, I appreciate tasks (whether it be gaming, writing, or programming) that end with a question and a sense that I don’t actually know very much. As noted in my last post, I raised a question about how our sense of place is different depending on its expression in terms of personification or metonymy. This question is not a result of the fact that I now have a jpeg of my findings, but it emerged in the process of constellating the texts, the code, and the maps.
And by constellate, as I roll my eyes at myself, I haphazardly threw together a series of abstractions so that I would eventually stare at my screen wondering how anything sensible actually managed to emerge.